Abstract: In order to solve the difficult problem of extracting degradation indicator of wind turbine bearings and the low prediction accuracy of remaining useful life（RUL）,a new prediction approach based on principal component analysis（PCA） and unscented particle filter（UPF） is proposed.The approach includes two steps:indicator extraction and RUL prediction.In the step of indicator extraction,original feature sets of real-time vibration signal of bearing are fused by using PCA approach.The degradation indicator which can reflect the deterioration trend of bearing is obtained.In the step of RUL prediction,a degradation model is constructed by fitting and analyzing the historical data of the bearing,the model parameters are updated by UPF to realize the degradation state tracking and prediction.Real test data of wind turbine bearings are involved to demonstrate the effectiveness of the proposed technique.Compared with PF,the proposed approach shows its superiority in particle degeneracy problem reduction and RUL prediction,which provides a reference for the health management and reliability evaluation of large wind turbines.